feedback on the use of ros in the infuse project
play

Feedback on the use of ROS in the InFuse project Ellon PAIVA MENDES - PowerPoint PPT Presentation

Feedback on the use of ROS in the InFuse project Ellon PAIVA MENDES Pierre NARVOR Journe ROS - robots mobiles terrestres emendes@laas.fr pnarvor@laas.fr Clermont-Ferrand 4 Juillet 2018 The InFuse project " Infusing Data Fusion in


  1. Feedback on the use of ROS in the InFuse project Ellon PAIVA MENDES Pierre NARVOR Journée ROS - robots mobiles terrestres emendes@laas.fr pnarvor@laas.fr Clermont-Ferrand – 4 Juillet 2018

  2. The InFuse project " Infusing Data Fusion in Space Robotics " • One of six projects of Space Robotics Technologies SRC (Horizon 2020) • Aims to develop of a Common Data Fusion Framework (CDFF) building block • To serve through all SRC upcoming activities • LAAS is only involved with the planetary rovers (not the orbital track) • CDFF development • Absolute map-based localization • Alternative perception techniques (hyper spectral cameras, lidars, etc)

  3. What is CDFF? • Common Data Fusion Framework • Defines a functional architecture to integrate the data fusion process • that is flexible and generic • with clear inner and outer interfaces • expose products (maps and positions) • and algorithm models (to allow their control)

  4. CDFF functional architecture (outputs) (inputs) CDFF CDFF products 1. Acquired data • IMU • Images • Point clouds • … Environment 2. Initial data & models models • Orbiter maps Core processes • Satellite models • … 3. Knowledge • Terramechanics • Dynamics Internal data structures • … Poses

  5. Project constraints • ESROCOS should be used as Robot Control Operational System (RCOS) • Space oriented RCOS being developed in a parallel project • Problem: ESROCOS is not ready yet! → We decided to use ROS instead. • Data exchanged between nodes should be described using ASN.1 • Interface description language for defining data structures that can be serialized and deserialized in a standard, cross-platform way. • We decided to user ASN.1 over ROS messages.

  6. ASN.1 Communication interface • ASN.1 (Abstract Syntax Notation One) widely used communication standard. • Exchanged types defined in high level .asn abstract description files. TASTE-BasicTypes DEFINITIONS ::= BEGIN -- Set of TASTE predefined basic types T-Int32 ::= INTEGER (-2147483648 .. 2147483647) T-UInt32 ::= INTEGER (0 .. 4294967295) T-Int8 ::= INTEGER (-128 .. 127) T-UInt8 ::= INTEGER (0 .. 255) T-Boolean ::= BOOLEAN END

  7. ASN.1 Communication interface • ASN.1 (Abstract Syntax Notation One) widely used communication standard. • Exchanged types defined in high level .asn abstract description files. • ASN1SCC : ESA’s ASN.1 compiler for safety-critical embedded systems. • .asn compiled into C files with ready to use serialization functions TASTE-BasicTypes DEFINITIONS ::= #ifndef GENERATED_ASN1SCC_TASTE_TYPES_H BEGIN #define GENERATED_ASN1SCC_TASTE_TYPES_H -- Set of TASTE predefined basic types typedef int T_Int32; T-Int32 ::= INTEGER (-2147483648 .. 2147483647) T-UInt32 ::= INTEGER (0 .. 4294967295) #define T_Int32_REQUIRED_BYTES_FOR_ENCODING 4 T-Int8 ::= INTEGER (-128 .. 127) #define T_Int32_REQUIRED_BITS_FOR_ENCODING 32 T-UInt8 ::= INTEGER (0 .. 255) #define T_Int32_REQUIRED_BYTES_FOR_ACN_ENCODING 4 T-Boolean ::= BOOLEAN #define T_Int32_REQUIRED_BITS_FOR_ACN_ENCODING 32 #define T_Int32_REQUIRED_BYTES_FOR_XER_ENCODING 39 END flag T_Int32_Encode(const T_Int32* val, BitStream* Serialization pBitStrm, int* pErrCode, flag bCheckConstraints); flag T_Int32_Decode(T_Int32* pVal, BitStream* pBitStrm, functions int* pErrCode); ...

  8. Separation between "core" and ASN.1 • Core functionalities are being written into libraries • Libraries are wrapped in/used by a "main" program that may be written in Genom3 + ROS or pure ROS • ASN.1 related code stays outside the libraries • ROS is used as implementation middleware

  9. ASN.1 and ROS • Libraries use any internal representation for data (pcl, eigen, etc) • Data is converted from/to an according ASN.1 type • The ASN.1 type is converted to/from bitstream using asn1scc encode/decode functions • Bitstream is stored in a ROS message for output to/input from the middleware

  10. A single ROS message type ● Only one type of ROS message is exchanged between the modules # ROS common header std_msgs/Header header # Identification of the type : ASN.1 name string type # Serialisation method : 0 (uPER), 1 (BER), 2 (XER) uint8 serialization_method # Buffer with ASN serialised data uint8[] data

  11. ASN.1 Tradeoffs • Pros • International widely used standard, mature technology • Simple text notation for type definition with physical encoding rules • Independent from programming languages • ASN1SCC compiler : free, open-source, ready to use • Cons • ASN.1 ROS message used are not human readable • Extra computation overhead (to be assessed) • Some encoding limitations (e.g. NaN or Inf encoding missing)

  12. Integration with docker • Part of the project may be supplied as dockers images • Allow easy use of software with closed source code • Dockers communicate through ROS using the same type of ROS message

  13. Visualization with rviz • Rviz plugin implemented to decode and display data

  14. LAAS Rovers (not on Mars) Mana Minnie

  15. LAAS Rovers Hardware • Segway rmp400 and rmp440 platforms. • Six-axis IMU • (accelerometer, gyrometer, magnetometer). • One axis Fiber-optic gyro. • (100Hz, drift of 1 deg/hour after correction). • RTK GPS. • (20Hz, cm accuracy => corrections provided by a nearby base).

  16. LAAS Rovers Hardware • Lidar rover (Mana): • Panoramic Lidar velodyne HDL64. ● (Scan rate 5-20Hz, vertical FOV -24/+2deg, 1.4M points per second) • Vision rover (Minnie): • Panoramic Lidar velodyne HDL32. ● (Scan rate 10Hz, vertical FOV -30/+10deg, 700k points per second) • 1 NavCam stereo bench on a PTU. • Automotive Lidar ● High resolution point cloud, FOV 110x90deg. • 2 HazCam fixed stereo benches (front and rear)

  17. LAAS Rovers System Diagram GPS Base + Robot Wifi Station Base GPS Sick Stereo Controller Receiver Lidar bench 1 2 Industrial computers: Velodyne Stereo Operator - DEM Building Lidar bench 2 - Lidar-based SLAM - Visual Odometry Hyper- Stereo - Position Manager spectral bench 3 - Data fusion + server camera - ROS bags PTU Controller

  18. Feedback from trials at CNES • First tests performed during the last two weeks • (in fact mostly integration work … ) • Data acquisition in ROS bags with joystick controlled robots • Needed to have some ASN.1 translator nodes here and there... • Single ROS message + ASN.1 helped integration with partners (Magellium) • Detected some latency/bandwidth problems • Less point clouds in the bags than expected (need to investigate why) • Acquired datasets still being analysed.

  19. Conclusion • Not an optimal solution • ROS being used only as communication layer + debugging • Still a work in progress … • Some questions: • Could we generate .msg from .asn files? • Should we use nodelets to reduce message passing? • Are we doing something very wrong here? :)

Recommend


More recommend